Videography and 360 degree rendering by Terence Rushin
Coda Datacenter
The Coda ecosystem is designed as a future-forward workplace enabling unique collaborations and partnerships between higher education and industry – both established and start-up – the overall theme is the nexus of CI, data sciences, and discovery focusing on HPC/HTC, “big data” storage and analytics capabilities, and leadership class network services including research and education (R&E) backbone access.
The Coda Data Center is the new home for PACE, offering increased room for expansion and allowing for much higher power density per rack — an important consideration for growth in areas such as machine learning and artificial intelligence that benefit from GPUs and other computational accelerators. The facility has expansion capability up to 9-10 megawatts of power in a modular fashion, enabling a dynamic response to evolving needs. Along with the Office Tower, the Data Center serves as a highly-instrumented Technology Showcase/Living Lab with access to complete operational data that is captured and available for researchers.
Owned and operated by DataBank, the Coda Data Center is a multi-story, state-of-the-art commercial data center facility. Three floors provide over 40,000 square feet of server space, with additional floors providing sup- porting mechanical and electrical space bringing the total facility up to 94,000 square feet. Initially, Georgia Tech has leased two megawatts of power: 1.5 megawatts in the Research hall, primarily allocated to PACE compute nodes; and another 0.5 megawatts in the Enterprise hall, used for critical PACE servers and storage, as well as OIT enterprise services. A number of sustainability technologies are being utilized in the Coda data center; waste heat from the data center is captured and provides heat to the office tower. Water recovered from an onsite fresh- water spring supplies 50+ gallons per minute of makeup water for the cooling systems, and temperatures within the data center are increased from the typical level, while staying within the applicable ASHRE standards. Combined, these technologies comprise a low PUE (Power Usage Effectiveness) design of 1.25 to deliver long-term operational cost savings. A research partnership with Georgia Power provides a self-contained power system nearby. This microgrid provides up to 1.5 megawatts of power to the data center using a variety of alternative technologies such as fuel cells, battery storage, and micro-turbines. The intent for the MicroGrid is to provide protection from brownouts during peak midtown grid utilization. The power feed will auto-switch diverse sources automatically, and this happens within the minimum flywheel UPS operation time of 15 seconds. The MicroGrid will auto start should both diverse sources fail or upon a Georgia Power directive. The facility is adaptable to incorporate future power generation technologies.
Currently, the Coda datacenter hosts PACE’s Hive, Phoenix, ICE clusters, and soon to be deployed CUI cluster for export controlled research. A brief overview of the current clusters in production are provided below.
Read more about Coda datacenter in the prior issue of the PACE Newsletter (page 7) and to read more about the great work accomplished by the Georgia Tech research community that was enabled by PACE resources, please see our prior issues of the PACE Newsletter.
As part of SC20 Student Tours, the Coda datacenter is one of the featured virtual tours for the high school and university students. Also, Coda datacenter virtual tour was featured in an HPCwire article.
Phoenix Supercomputer
The Phoenix cluster is Georgia Tech’s latest leading edge supercomputer that accelerates the computational research efforts of the Institute's diverse community in support of data-driven work in astrophysics, biology, health science, chemistry, materials and manufacturing, public policy, among other disciplines.
The Phoenix supercomputer is built in collaboration with Penguin Computing, Intel, Mellanox, and NVIDIA, and represents the largest single research computing investment by Georgia Tech in support of accelerating computational research. Phoenix is a heterogeneous cluster featuring Intel’s Cascade Lake processors and NVIDIA’s Volta and Quadro Pro graphics processing units, and consumes approximately 1 megawatt of power. The 34,440 cores deliver over 1.8 PF of performance based on the LINPACK Benchmark.
Please refer to this link about citing or acknowledging PACE if your research benefited from the Phoenix cluster.
Read more about Phoenix Supercomputer
Hive Supercomputer
The Hive Supercomputer is a $5.3 million HPC system that is enabling data-driven discovery in data science, computational astrophysics, biology, chemistry, and materials science at GT. Hive is a fulfillment of a recent $3.7M Major Research Instrumentation award to Georgia Tech from the National Science Foundation.
The Hive is built in collaboration with Penguin Computing, Intel, Mellanox, and NVIDIA. This heterogeneous cluster is comprised of 484 nodes featuring Intel’s Cascade Lake processors and NVIDIA’s VOLTA graphics processing cards. The 11,616 cores deliver 0.7 PF of performance based on the LINPACK Benchmark."
Please refer to this link about citing or acknowledging PACE if your research benefited from the Hive cluster.
Read more about Hive Supercomputer
ICE Clusters
In 2018, PACE initiated a project called “Instructional Cluster Environment (ICE)” to build instructional clusters in support of growing demand for coursework requiring HPC in accordance with Georgia Tech's strategic plan. These resources offer an educational environment that’s identical to production clusters, and expected to provide thousands of grad/undergrad students with ample opportunities to gain first-hand scientific computing experience including HPC and GPU programming each year. Furthermore, the entire PACE scientific software repository is made accessible to all ICE students, providing an education environment that mirrors production research clusters in every aspect.
In addition to credit-bearing courses, ICE enables PACE team to develop or host hands-on tutorials and workshops to help GT researchers and students improve their computational skills, in accordance with GT’s strategic plan that states “In addition to traditional classroom experiences, we will explore the possibility of enabling faculty, students and staff to interact in venues such as short courses and workshops where a shared culture of innovation, collaboration and leadership can evolve.”
